Kubeadapt

Metrics Glossary

Cost Metrics

Monthly Cost

Definition: Projected cost for entire month based on current resource allocation.

Formula:

text
1Monthly Cost = Hourly Cost × 24 hours × 30 days

Example:

text
1Hourly Cost: $0.384
2Monthly Cost: $0.384 × 24 × 30 = $276.48

Note: Uses 30 days as standard month (not actual calendar days).

Hourly Cost

Definition: Cost per hour for running resources.

Formula:

text
1Hourly Cost = (CPU Cost + Memory Cost + Storage Cost)

Components:

CPU Cost:

text
1CPU Cost = CPU Cores Requested × Cost per Core per Hour

Memory Cost:

text
1Memory Cost = Memory GB Requested × Cost per GB per Hour

Storage Cost:

text
1Storage Cost = Storage GB × Storage Cost per GB per Month / 720 hours

Example:

text
1Pod requesting:
2- 2 CPU cores @ $0.04/core/hour = $0.08/hour
3- 4 GB memory @ $0.005/GB/hour = $0.02/hour
4- 50 GB storage @ $0.08/GB/month = $0.0056/hour
5
6Hourly Cost = $0.08 + $0.02 + $0.0056 = $0.1056/hour
7Monthly Cost = $0.1056 × 720 = $76.03/month

Daily Cost

Definition: Cost per 24-hour period.

Formula:

text
1Daily Cost = Hourly Cost × 24

Example:

text
1Hourly Cost: $0.384
2Daily Cost: $0.384 × 24 = $9.22

Allocated Cost

Definition: Cost based on resource requests (what's reserved).

Formula:

text
1CPU Allocated Cost = CPU Request × CPU Unit Cost ($/core/hour)
2Memory Allocated Cost = Memory Request × Memory Unit Cost ($/GB/hour)
3
4Allocated Cost = CPU Allocated Cost + Memory Allocated Cost

Example:

text
1Pod requests: 500m CPU (0.5 cores), 1Gi memory
2
3Pricing (from cloud provider or Kubeadapt-opencost deployment):
4- CPU unit cost: $0.024/core/hour
5- Memory unit cost: $0.006/GB/hour
6
7CPU allocated cost = 0.5 cores × $0.024/core/hour = $0.012/hour
8Memory allocated cost = 1 GB × $0.006/GB/hour = $0.006/hour
9
10Allocated cost = $0.012 + $0.006 = $0.018/hour
11Monthly: $0.018 × 720 = $12.96/month

Actual Cost

Definition: Cost based on resource usage (what's actually used).

Formula:

text
1CPU Actual Cost = CPU Usage × CPU Unit Cost ($/core/hour)
2Memory Actual Cost = Memory Usage × Memory Unit Cost ($/GB/hour)
3
4Actual Cost = CPU Actual Cost + Memory Actual Cost

Example:

text
1Pod using: 200m CPU (0.2 cores), 600Mi memory (0.6 GB) - actual usage
2
3Pricing (same as above):
4- CPU unit cost: $0.024/core/hour
5- Memory unit cost: $0.006/GB/hour
6
7CPU actual cost = 0.2 cores × $0.024/core/hour = $0.0048/hour
8Memory actual cost = 0.6 GB × $0.006/GB/hour = $0.0036/hour
9
10Actual cost = $0.0048 + $0.0036 = $0.0084/hour
11Monthly: $0.0084 × 720 = $6.05/month

Note: For right-sizing recommendations, Kubeadapt uses percentile-based usage (P95 for CPU, P99 for memory) rather than instantaneous values to ensure reliable cost optimization.

Where used:

  • Efficiency calculations
  • Right-sizing recommendations
  • Waste identification

Idle Cost

Definition: Cost of unused/wasted resources.

Formula:

text
1Idle Cost = Allocated Cost - Actual Cost

Example:

text
1Allocated: $0.018/hour (based on requests)
2Actual: $0.0084/hour (based on usage)
3Idle: $0.018 - $0.0084 = $0.0096/hour wasted
4
5Monthly waste: $0.0096 × 720 = $6.91/month
6Waste percentage: ($0.0096 / $0.018) × 100% = 53% waste

Where used:

  • Waste reports
  • Over-provisioning detection
  • Optimization opportunities

Resource Metrics

CPU Request

Definition: CPU cores reserved for a pod (guarantee from Kubernetes).

Unit: Millicores (m) or cores

Example:

text
1500m = 0.5 cores = 50% of 1 CPU core
21000m = 1 core = 1 full CPU core
32000m = 2 cores

Where used:

  • Cost calculations (allocated)
  • Scheduling decisions
  • Resource quotas

CPU Usage

Definition: Actual CPU cores consumed by a pod.

Unit: Millicores (m) or cores

Measurement: Average over time period (typically 5-minute window)

For right-sizing: Kubeadapt analyzes historical usage patterns (30-day lookback) and uses P95 percentile for production workloads, P50 for non-production.

Example:

text
1Pod requests: 1000m (1 core)
2Pod using: 420m (42% of request)
3Utilization: 42%

Where used:

  • Efficiency calculations
  • Right-sizing recommendations
  • Performance monitoring

CPU Utilization

Definition: Percentage of requested CPU actually used.

Formula:

text
1CPU Utilization = (CPU Usage / CPU Request) × 100%

Example:

text
1CPU Request: 1000m
2CPU Usage: 420m
3CPU Utilization: (420 / 1000) × 100% = 42%

Where used:

  • Right-sizing recommendations
  • Performance analysis
  • Efficiency scoring

Memory Request

Definition: Memory reserved for a pod (guarantee from Kubernetes).

Unit: Bytes (Ki, Mi, Gi)

Example:

text
1512Mi = 536,870,912 bytes = ~512 MB
21Gi = 1,073,741,824 bytes = ~1.07 GB
32Gi = 2,147,483,648 bytes = ~2.14 GB

Where used:

  • Cost calculations
  • Resource quotas

Memory Usage

Definition: Actual memory consumed by a pod.

Unit: Bytes (Ki, Mi, Gi)

Measurement: Working set memory (RSS + cache)

For right-sizing: Kubeadapt analyzes historical usage patterns (30-day lookback) and uses P99 percentile for production workloads (conservative to prevent OOM), P50 for non-production.

Example:

text
1Pod requests: 2Gi
2Pod using: 1.4Gi (working set)
3Utilization: 70%

Where used:

  • Efficiency calculations
  • OOM risk detection
  • Right-sizing

Memory Utilization

Definition: Percentage of requested memory actually used.

Formula:

text
1Memory Utilization = (Memory Usage / Memory Request) × 100%

Example:

text
1Memory Request: 2Gi
2Memory Usage: 1.4Gi
3Memory Utilization: (1.4 / 2) × 100% = 70%

Where used:

  • Right-sizing recommendations
  • OOM prevention
  • Efficiency scoring

Storage Capacity

Definition: Total size of persistent volume claim.

Unit: Bytes (Gi, Ti)

Example:

text
1PVC size: 100Gi

Where used:

  • Storage cost calculations
  • Capacity planning

Storage Usage

Definition: Actual disk space used in volume.

Unit: Bytes (Gi, Ti)

Measurement: Used bytes on filesystem

Example:

text
1PVC capacity: 100Gi
2Used: 42Gi
3Utilization: 42%

Where used:

  • Storage efficiency
  • Growth trends
  • Resize recommendations

Storage Utilization

Definition: Percentage of volume capacity used.

Formula:

text
1Storage Utilization = (Storage Used / Storage Capacity) × 100%

Example:

text
1Capacity: 100Gi
2Used: 42Gi
3Utilization: 42%

Interpretation:

  • < 70%: Normal utilization
  • 70-85%: Good utilization

  • 85%: Consider expansion

Where used:

  • Resize recommendations
  • Waste detection
  • Capacity alerts

Efficiency Metrics

Resource Efficiency

Definition: How well resources are utilized (not wasted).

Formula:

text
1Resource Efficiency = (Observed Usage / Allocated Requests) × 100%

How Usage is Measured:

Kubeadapt uses percentile-based analysis to calculate observed usage:

  • CPU Usage: Based on P95 percentile of actual consumption
  • Memory Usage: Based on P99 percentile of actual consumption
  • Time-weighted: Recent data carries more weight than older data (30-day lookback with recency bias)

The percentile approach varies by environment:

  • Production workloads: P95 for CPU, P99 for memory (conservative)
  • Non-production workloads: P50 for both CPU and memory (aggressive cost optimization)

For combined CPU + Memory:

text
1Overall Efficiency = (CPU Efficiency + Memory Efficiency) / 2

Example:

text
1CPU: 420m usage (P95) / 1000m request = 42%
2Memory: 1.4Gi usage (P99) / 2Gi request = 70%
3Overall: (42% + 70%) / 2 = 56%

Target Ranges:

  • CPU: 60-75% efficiency (optimal balance between cost and performance)
  • Memory: 70-85% efficiency (optimal balance between cost and OOM risk)

Where used:

  • Optimization scoring
  • Cluster efficiency
  • Right-sizing recommendations

Bin-Packing Efficiency

Definition: How well pods fit on nodes (minimize fragmentation).

Formula:

text
1Bin-Packing Efficiency = (Used Resources / Allocatable Resources) × 100%

Cluster-level:

text
1Cluster Bin-Packing = Σ(All Pod Requests) / Σ(All Node Capacity)

Example:

text
1Cluster total allocatable: 100 cores, 400 GB
2Total pod requests: 68 cores, 288 GB
3CPU bin-packing: 68 / 100 = 68%
4Memory bin-packing: 288 / 400 = 72%
5Overall: (68% + 72%) / 2 = 70%

Interpretation:

  • < 60%: Poor utilization
  • 60-75%: Good utilization

  • 75%: Excellent utilization

Where used:

  • Node optimization
  • Cluster scaling decisions
  • Instance type recommendations

Cost Efficiency

Definition: How much of allocated cost is actually used.

Formula:

text
1Cost Efficiency = (Actual Cost / Allocated Cost) × 100%

Example:

text
1Allocated Cost: $0.018/hour (based on requests)
2Actual Cost: $0.0084/hour (based on usage)
3Cost Efficiency: (0.0084 / 0.018) × 100% = 47%
4
5Interpretation: 47% efficiency means 53% waste

Where used:

  • Savings calculations
  • Waste reports
  • Optimization ROI

Savings Metrics

Available Savings

Definition: Potential monthly cost reduction from optimization opportunities.

Formula:

text
1Available Savings = Σ (Optimization Opportunities)

Categories:

  • Over-provisioned workloads
  • Orphaned resources
  • Unused resources
  • Storage inefficiencies

Example:

text
1Over-provisioned: $1,100/month
2Orphaned: $180/month
3Unused: $40/month
4Total Available Savings: $1,320/month

Where used:

  • Optimization dashboard
  • ROI calculations
  • Executive reporting

Savings Percentage

Definition: How much cost can be reduced as percentage of total.

Formula:

text
1Savings % = (Available Savings / Total Cost) × 100%

Example:

text
1Available Savings: $1,320/month
2Total Cost: $7,240/month
3Savings %: (1,320 / 7,240) × 100% = 18.2%

Where used:

  • Optimization ROI
  • Comparison metrics
  • Executive dashboards

Implemented Savings

Definition: Actual savings achieved from implemented optimizations.

Formula:

text
1Implemented Savings = Baseline Cost - Optimized Cost
2Savings % = (Implemented Savings / Baseline Cost) × 100%

Example:

text
1Baseline Cost: $7,240/month
2Optimized Cost: $6,380/month
3Implemented Savings: $7,240 - $6,380 = $860/month (11.9%)

Where used:

  • Tracking optimization progress
  • ROI validation
  • Success metrics

Allocation Metrics

Allocated %

Definition: Percentage of cluster resources allocated to specific dimension (team, namespace, etc.).

Formula:

text
1Allocated % = (Dimension Cost / Total Cluster Cost) × 100%

Example:

text
1Team Cost: $3,240/month
2Total Cluster Cost: $7,240/month
3Allocated %: (3,240 / 7,240) × 100% = 44.8%

Where used:

  • Cost allocation
  • Chargeback reports
  • Team dashboards

Unallocated Cost

Definition: Cost of resources not attributed to any cost dimension.

Causes:

  • Resources missing required labels
  • System components (kube-system, monitoring)
  • Shared infrastructure

Formula:

text
1Unallocated Cost = Total Cost - Σ(Allocated Costs)
2Unallocated % = (Unallocated Cost / Total Cost) × 100%

Example:

text
1Total Cost: $7,240/month
2Allocated Costs: $6,890/month
3Unallocated: $7,240 - $6,890 = $350/month (4.8%)

Where used:

  • Cost allocation coverage
  • Label compliance tracking

Node Metrics

Node Cost

Definition: Monthly cost of running a single node.

Formula:

text
1Node Cost = Instance Hourly Rate × 24 × 30

Cost Breakdown:

For pod cost attribution, the bundled instance cost is split into separate CPU and memory components:

  • GCP: Real CPU and memory unit costs from Cloud Billing API
  • AWS/Azure: Calculated split using standard pricing ratios

Example:

text
1m5.2xlarge: $0.384/hour
2Node cost: $0.384 × 24 × 30 = $276.48/month

Where used:

  • Node-level cost tracking
  • Instance type comparison
  • Cluster cost breakdown

Node Utilization

Definition: Percentage of node capacity used by pods.

Formula:

text
1Node Utilization = Max(CPU Utilization, Memory Utilization)

Example:

text
1Node: 8 cores, 32 GB
2Pod requests: 5.4 cores, 22.1 GB
3
4CPU utilization: 5.4 / 8 = 67.5%
5Memory utilization: 22.1 / 32 = 69%
6
7Node utilization: Max(67.5%, 69%) = 69%

Note: Uses max because Kubernetes schedules based on whichever resource is most constrained.

Where used:

  • Node efficiency
  • Scaling decisions
  • Underutilization detection

Pod Density

Definition: Number of pods per node.

Formula:

text
1Pod Density = Pods Running / Node Capacity

Example:

text
1Node capacity: 58 pods max
2Pods running: 24
3Pod density: 24 / 58 = 41%

Where used:

  • Bin-packing analysis
  • Node sizing decisions

Workload Metrics

Workload Cost

Definition: Total cost of all pods in a workload (Deployment, StatefulSet, etc.).

Formula:

text
1Workload Cost = Pod Cost × Number of Replicas

Example:

text
1Pod Cost: $48/month
2Replicas: 4
3Workload Cost: $48 × 4 = $192/month

Where used:

  • Workload-level cost tracking
  • Scaling cost impact
  • Team cost allocation

Cost per Replica

Definition: Average cost of a single pod/replica.

Formula:

text
1Cost per Replica = Workload Cost / Number of Replicas

Example:

text
1Workload cost: $192/month
2Replicas: 4
3Cost per replica: $192 / 4 = $48/month

Where used:

  • Scaling decisions
  • HPA cost impact
  • Cost estimation

Forecast Metrics

Projected Monthly Cost

Definition: Estimated end-of-month cost based on current daily rate.

Formula:

text
1Projected Monthly = (Actual Spend / Days Elapsed) × Days in Month

Example:

text
1Nov 15 (15 days elapsed):
2Actual spend: $3,630
3Projected monthly: ($3,630 / 15) × 30 = $7,260

Where used:

  • Budget tracking
  • Month-end projections
  • Variance alerts

Run Rate

Definition: Annualized cost based on current monthly spend.

Formula:

text
1Annual Run Rate = Current Monthly Cost × 12

Example:

text
1November projected: $7,260
2Annual run rate: $7,260 × 12 = $87,120/year

Where used:

  • Annual budget planning
  • Executive reporting
  • Multi-year forecasting

Overhead

Definition: Infrastructure costs not directly attributable to workloads.

Components:

  • Control plane: EKS/GKE/AKS management fees (cloud-managed clusters only)
  • System processes: Kubelet, kube-proxy, system daemons (all clusters)
  • Idle node capacity: Unused CPU/memory after pod requests (scheduling waste)

Example:

text
1Total cluster: $7,240/month
2Workload costs: $6,890/month
3Overhead: $350/month (4.8%)

Where used:

  • True cost accounting
  • Workload cost allocation
  • Infrastructure efficiency

Calculation Notes

Time Periods

Standard periods:

  • Hourly: Near real-time tracking
  • Daily: 24 hours
  • Monthly: 30 days (720 hours)
  • Annual: 365 days

Note: Kubeadapt uses 30-day months (720 hours) for consistency.

Rounding

Display precision:

  • Costs: 2 decimal places ($276.48)
  • Percentages: 1 decimal place (45.7%)
  • Calculations: Full precision internally
Back to Docs
Edit this page